From 8912e0696af069de47646fdb8a9d9c4e086e88b3 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期日, 14 一月 2024 23:42:11 +0800
Subject: [PATCH] Resolve merge conflict
---
funasr/models/monotonic_aligner/model.py | 30 ++++++++++++++----------------
1 files changed, 14 insertions(+), 16 deletions(-)
diff --git a/funasr/models/monotonic_aligner/model.py b/funasr/models/monotonic_aligner/model.py
index ece319d..1b43c2f 100644
--- a/funasr/models/monotonic_aligner/model.py
+++ b/funasr/models/monotonic_aligner/model.py
@@ -5,17 +5,15 @@
from typing import Union, Dict, List, Tuple, Optional
from funasr.models.paraformer.cif_predictor import mae_loss
-from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
-from funasr.models.transformer.utils.nets_utils import make_pad_mask, pad_list
-from funasr.metrics.compute_acc import th_accuracy
from funasr.train_utils.device_funcs import force_gatherable
+from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
+from funasr.models.transformer.utils.nets_utils import make_pad_mask
from funasr.utils.timestamp_tools import ts_prediction_lfr6_standard
from funasr.utils import postprocess_utils
from funasr.utils.datadir_writer import DatadirWriter
from funasr.register import tables
from funasr.models.ctc.ctc import CTC
-from funasr.utils.load_utils import load_audio_and_text_image_video, extract_fbank
-
+from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
@tables.register("model_classes", "monotonicaligner")
@@ -25,7 +23,6 @@
Achieving timestamp prediction while recognizing with non-autoregressive end-to-end ASR model
https://arxiv.org/abs/2301.12343
"""
-
def __init__(
self,
input_size: int = 80,
@@ -41,7 +38,6 @@
length_normalized_loss: bool = False,
**kwargs,
):
-
super().__init__()
if specaug is not None:
@@ -155,11 +151,10 @@
frontend=None,
**kwargs,
):
-
meta_data = {}
# extract fbank feats
time1 = time.perf_counter()
- audio_list, text_token_int_list = load_audio_and_text_image_video(data_in,
+ audio_list, text_token_int_list = load_audio_text_image_video(data_in,
fs=frontend.fs,
audio_fs=kwargs.get("fs", 16000),
data_type=kwargs.get("data_type", "sound"),
@@ -171,7 +166,8 @@
meta_data["extract_feat"] = f"{time3 - time2:0.3f}"
meta_data["batch_data_time"] = speech_lengths.sum().item() * frontend.frame_shift * frontend.lfr_n / 1000
- speech.to(device=kwargs["device"]), speech_lengths.to(device=kwargs["device"])
+ speech = speech.to(device=kwargs["device"])
+ speech_lengths = speech_lengths.to(device=kwargs["device"])
# Encoder
encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)
@@ -190,13 +186,15 @@
timestamp_str, timestamp = ts_prediction_lfr6_standard(us_alpha[:encoder_out_lens[i] * 3],
us_peak[:encoder_out_lens[i] * 3],
copy.copy(token))
- text_postprocessed, time_stamp_postprocessed, word_lists = postprocess_utils.sentence_postprocess(
- token, timestamp)
+ text_postprocessed, time_stamp_postprocessed, _ = postprocess_utils.sentence_postprocess(token, timestamp)
result_i = {"key": key[i], "text": text_postprocessed,
"timestamp": time_stamp_postprocessed,
- }
- # ibest_writer["token"][key[i]] = " ".join(token)
- ibest_writer["timestamp_list"][key[i]] = time_stamp_postprocessed
- ibest_writer["timestamp_str"][key[i]] = timestamp_str
+ }
results.append(result_i)
+
+ if ibest_writer:
+ # ibest_writer["token"][key[i]] = " ".join(token)
+ ibest_writer["timestamp_list"][key[i]] = time_stamp_postprocessed
+ ibest_writer["timestamp_str"][key[i]] = timestamp_str
+
return results, meta_data
\ No newline at end of file
--
Gitblit v1.9.1